Prediction and Mapping of Understorey Vegetation Coverage and Change Analysis in Fennoscandian Ecosystem Using Satellite Remote Sensing Data

Understory vegetation (USV) is a crucial part of forest ecosystems, playing multiple roles such as shaping forest structure, cycling nutrients, and influencing species diversity.

It serves as a haven for wildlife, offering food and shelter, and provides valuable resources like wild mushrooms and berries.

Understanding the characteristics and distribution of USV across different landscapes and seasons is crucial for managing forests effectively.

This knowledge aids in better decision-making for activities such as forest regeneration, conservation efforts, berry picking, and assessing the risk of forest fires.

Despite the importance of USV, accurately mapping its coverage has been challenging, primarily due to the obstructed view from the forest canopy.

In our project, we aim to improve the possibility of mapping USV by analysing various spectral and structural variables extracted from satellite imagery.

These variables, including vegetation height and spectral signatures, will help us map the percentage of ground surface covered by USV.

We will combine this satellite data with field observations from the Swedish National Forest Inventory and with other environmental data, such as soil moisture and terrain maps, to improve mapping accuracy.

To achieve this, we will develop advanced artificial intelligence algorithms capable of predicting USV species and coverage percentage.

These algorithms will be compared with traditional parametric models to assess their performance of predicting USV species and coverage percentage.

Additionally, we will validate our models using remote sensing data acquired from aerial platforms and assess their accuracy across different canopy cover classes. Furthermore, we will explore the combination of data from multiple satellite sensors to enhance the predictions.

By combining spectral information available from optical imagery with structural data derived from synthetic aperture radar (SAR) data, we aim to improve the accuracy of USV species and coverage predictions.

Finally, we will conduct a time-series analysis of satellite images from two time periods to understand how the distribution and ground coverage of USV change over time.

The results from the project will be integrated into the national forest data product, Skogliga Grunddata, and shared as open data. It will also serve as an initiation for creating a new benchmark database for USV data.